Fluorescence-based temperature control for polymerase chain reaction

Analytical Biochemistry 448 (2014) 75–81 Contents lists available at ScienceDirect Analytical Biochemistry journal homepage: www.elsevier.com/locate/yabio Fluorescence-based temperature control for polymerase chain reaction q Lindsay N. Sanford a, Carl T. Wittwer b,⇑ a Department of Bioengineering, University of Utah, Salt Lake City, UT 84112, USA b Department of Pathology, University of Utah Health Sciences Center, Salt Lake City, UT 84112, USA a r t i c l e i n f o a b s t r a c t Article history: The ability to accurately monitor solution temperature is important for the polymerase chain reaction Received 14 October 2013 (PCR). Robust amplification during PCR is contingent on the solution reaching denaturation and annealing Received in revised form 20 November 2013 temperatures. By correlating temperature to the fluorescence of a passive dye, noninvasive monitoring of Accepted 22 November 2013 solution temperatures is possible. The temperature sensitivity of 22 fluorescent dyes was assessed. Emis- Available online 28 November 2013 sion spectra were monitored and the change in fluorescence between 45 and 95 °C was quantified. Seven dyes decreased in intensity as the temperature increased, and 15 were variable depending on the excita- Keywords: tion wavelength. Sulforhodamine B (monosodium salt) exhibited a fold change in fluorescence of 2.85. Polymerase chain reaction (PCR) Fluorescence Faster PCR minimizes cycling times and improves turnaround time, throughput, and specificity. If tem- Temperature monitoring perature measurements are accurate, no holding period is required even at rapid speeds. A custom instru- High-resolution melting ment using fluorescence-based temperature monitoring with dynamic feedback control for temperature cycling amplified a fragment surrounding rs917118 from genomic DNA in 3 min and 45 s using 35 cycles, allowing subsequent genotyping by high-resolution melting analysis. Gold-standard thermocouple read- ings and fluorescence-based temperature differences were 0.29 ± 0.17 and 0.96 ± 0.26 °C at annealing and denaturation, respectively. This new method for temperature cycling may allow faster speeds for PCR than currently considered possible. Ó 2013 Elsevier Inc. All rights reserved. The ability to accurately monitor and control solution tempera- allow for faster cycling. Furthermore, the ability to dynamically ture during PCR1 and melting analysis is crucial for successful ampli- control thermal cycling based on actual solution (and not external) fication and analysis of high-resolution melting curves. Due to temperature is enabled. The idea of utilizing fluorescence to con- practical limitations, temperature measurements are typically made trol PCR temperature cycling was first suggested in 1994 [5], and externally to the sample solution. This produces significant solution- some progress has been made toward that goal. Fluorescence mon- instrument temperature mismatches [1] which are exacerbated dur- itoring has been used to control extension times during PCR [6]. ing rapid temperature transitions. One proposed solution for amelio- This allowed for heating (toward denaturation) to begin as soon rating solution-instrument temperature discrepancies is to correlate as extension was complete, instead of waiting for a prespecified the fluorescence of a passive dye (one that does not interact with time interval to pass. However, this approach used a fluorescent DNA) to solution temperature for noninvasive monitoring in real dye that also interacts with DNA, potentially confounding the fluo- time. This approach has been successfully demonstrated in harsh rescent signals for temperature measurement and real-time PCR environment flow-field applications [2,3], to make temperature monitoring. maps of microfluidic systems [4], and more recently, in commercial An alternative approach is to use a temperature-sensitive dye PCR instruments [1] to assess the differences in temperature that does not interact with DNA (and that does not inhibit the recorded by the instrument and actual PCR solution temperatures. PCR), but responds to changes in temperature with altered emis- The ability to noninvasively monitor solution temperatures sion intensity. For instance, the fluorescence of a temperature-sen- during PCR may provide more accurate sample temperatures and sitive dye has been used to adjust the power of a laser used to heat PCR mixtures from annealing to denaturation temperatures in a q droplet-based PCR system [7] to demonstrate fluorescence-based Partial support for this research was provided by BioFire Diagnostics and Canon US Life Sciences. heating control. However, complete thermal cycling control during ⇑ Corresponding author. Address: Department of Pathology, University of Utah PCR using fluorescence measurements requires the ability to con- School of Medicine, 50 N. Medical Dr., Salt Lake City, UT 84132, USA. Fax: +1 801 trol all parts of PCR including heating, cooling, and holding times. 581 6001. We begin with an examination of 22 alternative dyes for fluo- E-mail address: [email protected] (C.T. Wittwer). 1 rescence-based temperature monitoring, evaluating temperature Abbreviation used: PCR, polymerase chain reaction. 0003-2697/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ab.2013.11.027 76 Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 stability, fluorescence quenching, and the magnitude of the cali- imately 20 min. Dye concentrations and excitation wavelengths bration constant that relates temperature to fluorescence. Then, are shown in Table 1. After the sample temperature reached we select one of these dyes to noninvasively monitor solution tem- 95 °C, samples were cooled to 45 °C (using passive cooling). The peratures, and finally use the fluorescence-based temperature to spectrum was taken at 45 °C before and after heating and the control PCR cycling with simultaneous real-time amplification out- intensity of peak spectral bands compared. put and subsequent melting analysis for genotyping single nucleo- tide variants. Fluorescence-based and thermocouple temperature comparisons during melting Materials and methods Six dyes exhibiting the greatest sensitivity to temperature were Fluorescent dye temperature sensitivity further examined. Dye concentrations and excitation wavelengths are detailed in Table 2. Calibration constants (see Table 2) for each The temperature sensitivity of 22 fluorescent dyes was exam- dye were calculated from temperature and fluorescence data ac- ined on a custom multicolor fluorimeter with xenon excitation, quired from 45 to 95 °C in 10 °C increments. Using a second sample spectral dispersion (405–590 nm) on a grating, and focusing onto with a 30 lL final volume (25 lL sample with 5 lL oil overlay), an a fiber optic (delta RAM, Photon Technology International). The fi- initial holding period at 50 °C was used to determine reference ber optic illuminated the end of a glass capillary (LightCycler, temperature and intensity values. The sample was then heated to Roche Applied Science) placed within a heating unit (HR-1, BioFire 95 °C at an approximate rate of 0.05 °C/s. Fluorescence-based solu- Diagnostics). Fluorescent emission was collected by another fiber tion temperatures were calculated for both single-dye/single-color optic at a right angle to the capillary, delivering light onto a CCD and single-dye/two-color configurations using sulforhodamine B spectrometer (DV420-OE, Andor Technology), and collecting 1024 (acid form). Solution temperatures using the single-dye/two-color bins between 400 and 850 nm. A J-type miniature thermocouple method were calculated as described previously [8]. Fluorescence- (5SRTC-TT-J-40-36, Omega) was inserted into the sample capillary based temperatures were compared to sample thermocouple for physical temperature measurements. readings. Fluorescent dyes were tested in a ‘‘mock’’ (no polymerase) PCR solution consisting of 50 mM Tris (pH 8.3), 2 mM MgCl2, 0.2 mM Instrumentation for fluorescence-based temperature cycling control each deoxynucleotide triphosphate (Roche), 500 lg/ml bovine ser- um albumin (Sigma), and 0.08% (v/v) glycerol. Final concentrations Fluorescence-based temperatures were calculated by monitor- and excitation wavelengths for each dye are listed in Supplemental ing sulforhodamine B (monosodium salt, 600 lM, Sigma–Aldrich) Table 1. Spectra for temperature-sensitivity profiles of 22 dyes fluorescence. A LightCycler 24 (BioFire Diagnostics) was modified were recorded at 4 excitation wavelengths selected from 405, to accept voltage inputs from both standard thermocouples and 455, 470, 490, and 530 nm. Twenty-five microliter sample volumes fluorescence-based equivalents. Additional hardware is listed in were heated from 45 to 95 °C (in discrete 10 °C increments, with Supplemental Table 2. A (5SRTC-TT-J-40-36) thermocouple was 100 points averaged at each step). The ramp rate between steps placed in an adjacent capillary for the initial calibration to deter- during heating was approximately 0.4 °C/s. The entire emission mine reference values and for comparison to fluorescence-based spectrum was recorded and the fold change in fluorescence temperatures during cycling. (45–95 °C) calculated over selected bandwidths. PCR Fluorescent dye degradation Three forensic single-nucleotide variants [9] were amplified The system described above was used to assess the amount of using the following primers: rs876724 (50 -CCACTGCACTGAAGTA- dye degradation after a single heating and cooling cycle of approx- TAAGT-30 and 50 -TTAGCAGAGTGTGACAAAAAA-30 ), rs917118 (50 Table 1 Dye and spectral band selection. Dye Concentration (lM) Excitation wavelength (nm) Degradation Temperature–sensitive Temperature–insensitive (Fold-change)a band (nm) band (nm) Ethyl eosin 800 455 0.76 535–600 490–515 Ethyl eosin 800 470 0.83 535–590 490–515 Merocyanine 540 40 470 0.41 580–650 500–525 Merocyanine 540 40 490 0.54 575–625 525–550 Merocyanine 540 40 530 0.53 575–700 530–550 Rhodamine B 4 470 0.99 570–610 525–550 Rhodamine B 4 490 0.98 570–610 525–550 Rhodamine B 4 530 0.98 560–590 514–540 Snarf-1 160 455 0.96 630–680 500–525 or 580–600 Snarf-1 160 470 0.97 630–680 500–525 or 560–600 Snarf-1 160 490 0.89 625–675 525–575 Sulforhodamine B (acid form) 8 455 0.85 575–610 500–550 Sulforhodamine B (acid form) 8 470 0.93 575–620 500–550 Sulforhodamine B (acid form) 8 490 0.95 575–620 525–550 Sulforhodamine B (acid form) 8 530 0.86 575–620 525–550 Sulforhodamine B (monosodium salt) 4000 470 1.0 575–610 516–545 Sulforhodamine B (monosodium salt) 2400 490 0.98 575–610 516–545 Sulforhodamine B (monosodium salt) 2400 530 0.97 575–610 516–545 a Samples were heated from 45 to 95 °C and then cooled to 45 °C. Values were calculated at the emission peak wavelength. The fluorescence at the second measurement was divided by the fluorescence at the first measurement. Thus, a value of 1 indicates that no degradation occurred.  Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 77 Table 2 Fluorescence-based and thermocouple temperature comparisons during melting using one- and two-color bands. Dye Concentration Excitation Utilized Emission (nm) Calibration Temperature difference (lM) wavelength bands constant (Fluorescence–Thermocouple) (nm) Maximum (°C)a Mean ± SD (°C)b Ethyl eosin 800 470 1 550–625 1445 9.8 2.2 ± 4.9 Merocyanine 540 40 490 1 575–625 4267 21.3 7.1 ± 3.7 Rhodamine B 4 490 1 574–610 2551 1.1 0.1 ± 0.5 Snarf-1 160 490 1 640–675 1712 5.4 4.1 ± 1.5 Sulforhodamine B (acid form) 80 490 1 574–610 2570 0.6 0.4 ± 0.1 Sulforhodamine B (acid form) 80 490 2 525–550 and 1856 0.3 0.1 ± 0.1 574–610 Sulforhodamine B 4000 490 1 574–610 2040 2.4 1.2 ± 0.9 (monosodium salt) a The absolute temperature difference between fluorescence-based and thermocouple temperatures. b Thermocouple temperatures are subtracted from fluorescence-based temperatures. AAGATGGAGTCAACATTTTACAAG-30 and 50 -GATGACTGAGGTCAAC- GAG-30 ), and rs763869 (50 -AGGATGTTTGTTTATATTATTTCTAACT- CA-30 and 50 -CTACTCCCTCATAATGTAATGC-30 ). Each 10 lL reaction contained 50 mM Tris–HCl (pH 8.3), 2.0 mM Mg2+, 200 lM each deoxynucleoside triphosphate, 1X LCGreen+ (BioFire Diagnostics), 0.4 U KlenTaq1 (Ab Peptides) with 64 ng anti-Taq antibody (eEn- zyme), 0.5 lg BSA, 0.5 lM primers, and 50 ng genomic DNA. After an initial denaturation at 95 °C for 1 min, amplification conditions on the modified LightCycler were 85 °C for 10 s and 60 °C for 10 s for 35 cycles. To further demonstrate the utility of fluorescence-based tem- perature control for use with very rapid protocols, the rs917118 target was amplified using ‘‘0’’s hold times. The use of ‘‘0’’s hold times at annealing was enabled by increasing the concentration of primers by 5-fold (final concentration 2.5 lM). After an initial denaturation at 95 °C for 1 min, 35 cycles of 83 °C for ‘‘0’’s and Fig.1. Temperature-sensitivity profile for (2.4 mM) sulforhodamine B (monoso- 61 °C for ‘‘0’’s were performed. In all cases thermal cycling was dium salt) excited at 530 nm. Twenty-five microliter sample volumes were held at 45 (black) to 95 °C (light gray) in 10 °C intervals. Regions of high-temperature controlled by fluorescence-based temperature measurements. sensitivity (near 580 nm) are clearly discernible from regions of low-temperature sensitivity (near 525 nm). Melting analysis Ten microliter samples were melted on a HR-1 high-resolution were excluded from further consideration. Ethyl eosin, merocya- instrument (BioFire Diagnostics) at a rate of 0.3 °C/s. The instru- nine 540, rhodamine B, snarf-1, sulforhodamine B (acid form), ment was modified to accept LightCycler 24 capillaries. Melting and sulforhodamine B (monosodium salt) were studied further. curves were analyzed using the quantum method of background To better identify regions of the emission spectrum that should removal [10] and normalized between 0 and 100%. Savitsky–Golay be monitored to obtain optimal temperature sensitivity, spectral polynomials [11] were used to calculate derivative plots so that data were displayed as a ratio (emission intensity at 45 °C/emis- genotypes could be better visualized. sion intensity at 95 °C) on a log scale across wavelength (Fig. 2). Regions of low-temperature sensitivity will be close to zero while Results regions of high-temperature sensitivity have peaks or valleys fur- thest away from zero. Optimal temperature-sensitive and -insensi- The temperature sensitivity of sulforhodamine B (monosodium tive bands for each dye were identified and are listed in Table 1. salt), excited at 530 nm, is shown in Fig. 1. For each of 22 dyes, the Dye degradation caused a decrease in fluorescence after a single fold change in fluorescence during heating from 45 to 95 °C is heating and cooling cycle (Table 1). Sulforhodamine B (monoso- shown in Supplemental Table 1, along with general trends in fluo- dium salt), excited at 470 nm, had a value of 1, indicating that no rescence (i.e., increased or decreased intensity levels) that occurred degradation occurred. Values for ethyl eosin (excited at 455 nm) during heating. The average fold change in fluorescence for all dyes and merocyanine 540 (excited at 470 nm) were 0.76 and 0.41, was 0.74 (±0.45). Eosin B, excited at 530 nm, exhibited the greatest respectively. fold change in fluorescence (fluorescence decreased by a factor of Using the most temperature-sensitive wavelengths, fluores- 12.48), although differences in fluorescence levels at 85 and cence-based solution temperatures were determined during melt- 95 °C were limited. Changes in spectral intensity for each dye are ing. The maximum and mean (+/–SD) temperature differences compiled in an online database (https://www.dna.utah.edu/ between thermocouple and fluorescence-based temperatures are dyedb). shown in Table 2. The best performing single-band result was pro- Dyes exhibiting at least a 2-fold change in fluorescence were se- duced by sulforhodamine B (acid form, excited at 490 nm), with a lected for more in-depth examination. Dyes that exhibited variable maximum temperature difference of 0.6 ± 0.1 °C. However, this trends in fluorescence with increasing temperature or minimal dif- result was improved to 0.3 ± 0.07 °C when a second color ferences (<5%) in fluorescence at high temperatures (85 to 95 °C) band, insensitive to temperature, was used for normalization. A 78 Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 color band ratio better track thermocouple readings than a single color over the melting transition. Because sulforhodamine B (monosodium salt) exhibited the strongest temperature sensitivity as well as the most repeatable fluorescence after heating and cooling, it was used for thermal cycling control. A schematic diagram of fluorescence-based tem- perature cycling control is shown in Fig. 4. A representative time–temperature trace showing control by fluorescence-based temperatures with comparison to thermocouple-based tempera- tures is shown in Fig. 5. Three cycles are shown in Fig. 5 while Supplemental Fig. 1 shows all 35 cycles. When traces were time-aligned, the average temperature difference between fluorescence-based and thermocouple-based temperatures for all cycles was 0.5 °C (±0.4). These differences were 0.29 ± 0.2 and 0.96 ± 0.3 °C at annealing and denaturation extremes, respectively. Three forensic single nucleotide variants were amplified utiliz- Fig.2. Temperature sensitivity of sulforhodamine B (monosodium salt) between 45 and 95 °C with excitation at 490 nm. Spectral data used for the ratio calculation on ing fluorescence-based thermal cycling control with 10 s holds the Y-axis were generated using 25 lL volumes. When data are displayed in this resulting in quantification cycles (Cq) values of 24 (data not way, temperature-insensitive regions are near 0.0 and temperature-sensitive shown). Melting analysis identified all genotypes (see Supplemental regions are the peaks and valleys of greatest distance away from 0.0. Fig. 2), with both homozygotes and heterozygotes clearly distinguishable. The rs917118 target was also used to demon- strate successful amplification under ‘‘0’’s holding times, requiring only 3 min and 45 s to complete 35 cycles (Fig. 6A). While the Cq increased to 28, all 3 genotypes were clearly identified by high-res- olution melting of small amplicons (Fig. 6B). Discussion Several temperature-sensitive fluorescent dyes have been iden- tified for potential use in temperature monitoring and control of PCR. In the most straightforward implementation, a dedicated sample containing the fluorescent temperature-sensitive dye is monitored to control the cycling of other samples that are ampli- fied by PCR. More advanced schemes involve adding the fluores- cent dye directly to the PCR mixture so that real-time amplification and temperature monitoring for cycling control may be conducted simultaneously for each sample or zone. This is attractive because temperature validation can be performed on Fig.3. Fluorescence-based temperatures using one- and two-color bands compared each sample, although color compensation may be required due to thermocouple readings during melting. A 25 lL sample containing sulforhoda- to spectral overlap [12]. Sulforhodamine B is strongly tempera- mine B (acid form, 80 lM) was heated to 95 °C at an approximate rate of 0.05 °C/s. ture-sensitive and does not inhibit PCR or show much thermal deg- Fluorescence-based temperature data using one-(dark gray plus sign) and two- (light gray) color bands are compared to thermocouple data (black). Differences in radation [1]. fluorescence-based temperature and thermocouple readings were reduced with the Different methods have been used to correlate fluorescence use of two-color bands. The absolute temperature differences between the emission with solution temperature. In the most straightforward thermocouple and the single-color temperature (dark gray plus sign) and the approach (termed single-dye/single-color), a single dye is excited thermocouple (TC) and the two-color temperature (light gray) are displayed within at a specific wavelength and changes in emission intensity are the inset. Average differences between the thermocouple and the temperatures derived using single-color fluorescence were 0.39 °C. Average temperature differ- monitored in a single spectral band [2,3,13,14]. Better stability ences were reduced to 0.1 °C when two-color fluorescence was used. can be obtained with ratios of two (single-dye/two-color) [8,15] or even three (single-dye/three-color) [16,17] spectral bands. The comparison of thermocouple data and fluorescence-based temper- fluorescence intensity of a spectral band that is sensitive to tem- atures calculated using one- and two-color bands is shown in perature is normalized by the intensity of a second spectral band Fig. 3. Fluorescence-based temperatures calculated using a two- that is insensitive to temperature. By lowering noise from Fig.4. A schematic diagram of fluorescence-based temperature control. The fluorescent dye is excited by an LED light source, with emission intensity detected by a PMT. Excitation and emission pathways are filtered through the optical block. (A) In an initial calibration, reference intensity (voltages) and temperature measurements (via a thermocouple) are recorded and used to calculated fluorescence-based temperatures. (B) These temperatures are then used to control thermal cycling during PCR.  Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 79 dye/single-color). The ratio is then taken between the emission intensities measured under excitation at 2 different wavelengths. [18]. Other approaches achieve normalization with two dyes (two-dye/two-color) configurations. A temperature-sensitive dye is normalized using the signal from another dye with similar [19,20] or opposing temperature sensitivity [21,22]. While progress in the sophistication of fluorescent systems to improve temperature accuracy has occurred, the dyes for deter- mining fluorescence-based temperatures have largely remained the same. Reviewing 25 studies conducted from 1993 to 2012 using a single dye (measured on one, two, or three spectral bands) for fluorescence–temperature correlations, 68% utilized rhodamine B and 12% used sulforhodamine B. When a single-dye/single-color configuration was used, rhodamine B was selected 100% of the time. Two-dye/two-color configurations were similarly limited. When a temperature-sensitive dye was normalized using a dye Fig.5. Time–temperature trace produced with fluorescence-based temperature of similar temperature sensitivity, rhodamine B or sulforhodamine control (gray) compared to thermocouple-based temperatures (black). Fluores- B coupled with fluorescein or a fluorescein derivative was used cence-based temperatures were smoothed using Savitsky–Golay polynomials and thermocouple readings were smoothed with a passive RC circuit. The thermocouple 100% of the time. In studies using two dyes of opposing tempera- trace lags behind fluorescence-based readings by approximately 1.3 s because of ture sensitivity, a combination of rhodamine B normalized by rho- the RC delay. All 35 cycles are shown in Supplemental Fig. 1. For denaturation, the damine 110 was used 71% of the time. Studies examining more temperature peaks for fluorescence- and thermocouple-based temperatures were diverse dyes [18,23] focus on fluorescent pH indicators. 84.1+/–0.5 and 85.0+/–0.6 °C, respectively, with their mean difference 0.96+/– 0.26 °C. For annealing, the temperature troughs for fluorescence- and thermocou- In this study, those dyes exhibiting the greatest temperature ple-based temperatures were 63.3+/–0.3 and 63.6+/–0.4 °C, respectively, with their sensitivity during heating were given primary focus. However, mean difference 0.29+/–0.17 °C. many two-dye/two-color approaches achieve normalization though the use of temperature-insensitive dyes. Supplemental Ta- ble 1 identifies dyes that exhibit both high- and low-temperature sensitivities. Suitable passive fluorescent dyes for PCR must be thermally sta- ble for repeated heating and cooling without appreciable thermal degradation. Of the six dyes examined, ethyl eosin and merocya- nine 540 were substantially degraded (Table 1). Optimal spectral regions for both temperature sensitivity and insensitivity were identified (Table 1). Because low-temperature sensitivity often occurred around 500 nm, excitation wavelengths were typically set below this at 470 or 490 nm. When melting anal- ysis was performed, merocyanine 540 and ethyl eosin had the greatest mismatch between thermocouple and fluorescence-based temperatures (21.3 and 9.8 °C, respectively), further eliminating them from consideration (Table 2). Sulforhodamine B (monoso- dium salt, purity 75%), sulforhodamine B (acid form, purity 95%), and rhodamine B matched within 3 °C, supporting their use as dyes for temperature monitoring. Additionally, photodegradation of the fluorescent dye selected for use in determining fluorescence-based temperature is also of concern. On the custom instrumentation described above (LED excitation at 530 nm, 100% intensity) dye fluorescence decreased by 1.63% (an average of 0.05%/min) during a 30 min holding period. Given that we were able to obtain sufficient signal levels using 25% LED intensity and that the entire PCR protocol was performed in under 4 min, concerns regarding the occurrence of photodegrada- tion at levels that could potentially impact the accuracy of fluores- cence-based temperature calculations are minimized. Although the potential for photodegradation should be carefully considered, the excitation wavelength, intensity, and frequency of illumination can Fig.6. Amplification using ‘‘0’’s hold times with fluorescence-based temperature control. Thirty-five cycles of 83 °C for ‘‘0’’s and 61 °C for ‘‘0’’s were used to amplify be experimentally controlled to minimize such detrimental effects, rs917118 (C > T) with a Cq of 28. (A) Real-time amplification curve. (B) Melting and prior work with real-time PCR suggests that it is not a major results analyzed using the quantum method of background removal. All 3 concern [1]. genotypes [wild-type (light gray), variant (dark gray), and heterozygotes (black)] Methods using one or two emission bands were examined for are clearly distinguishable. sulforhodamine B (acid form). Increased temperature accuracy was achieved with two spectral bands. The maximum temperature difference between the thermocouple and the two-color analysis excitation light intensity, sample volumes, or optical path varia- was 0.3 °C, compared to 0.6 °C with a single color. Implementation tion, ratio measurement do improve temperature accuracy. of a fluorescence ratio decreased the error by about a factor of 2. A single dye (typically one that is sensitive to pH) can be excited Using the thermocouple temperatures as a reference, single-color at two different excitation wavelengths (dual excitation single- and two-color absolute temperature differences are shown by the 80 Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 inset to Fig. 3. The average error with one color is 0.39 °C. Although Acknowledgments the average two-color error is 0.1, variation up to 0.3 °C is present. This should not be interpreted as a general limitation of fluores- The authors extend their sincere thanks to Derek David for his cence-based temperature measurement, but instead as a demon- assistance in modifying hardware (LightCycler24, BioFire Diagnos- stration of the need for experimental enhancements that may tics) for fluorescence-based temperature measurements. The include more stable excitation sources, elimination of the thermo- authors also thank Keith Carney and Olewole Elenatoba-Johnson couple from the sample solution (which could potentially interfere for their work on the multicolor instrument, including program- with fluorescence acquisition), and/or the use of smoothing algo- ming of its LabView software. The authors thank Zachary Dwight rithms to aid in reducing variation in emission intensity not pro- for his work constructing the fluorescent dye temperature-sensi- duced by temperature variations. tivity database. Finally, we thank Jana Kent for use of the assays Single nucleotide variants were successfully amplified using used to demonstrate successful PCR with fluorescence-based tem- fluorescence-based temperature control. Cq values in the 20’s sug- perature control. gest efficient amplification, even when minimal holding times of 10 s or even ‘‘0’’ s are used. Zero second holding times required higher primer concentrations to speed the annealing process. Appendix A. Supplementary data Accurate temperature cycling allows cycle times to be minimized, as equilibrium holding times (imposed to allow the solution tem- Supplementary data associated with this article can be found, in perature to ‘‘catch up’’ to instrument readings) can be reduced or the online version, at http://dx.doi.org/10.1016/j.ab.2013.11.027. eliminated altogether. In a 35 cycle comparison of fluorescence- and thermocouple- based temperature cycling control (Fig. 5 and Supplemental Fig. 1), References the average difference between temperature measurements was [1] L. Sanford, C. Wittwer, Monitoring temperature with fluorescence during real- 0.5 ± 0.4 °C. When differences at annealing and denaturation were time PCR and melting analysis, Anal. Biochem. 434 (2013) 26–33. considered, these differences were 0.3 ± 0.2 and 0.9 ± 0.3 °C, respec- [2] P. Lavieille, F. Lemoine, G. Lavergne, F. Virepinte, M. Lebouché, Temperature tively. Thus, the fluorescence-based temperature tracks well with measurements on droplets in monodisperse stream using laser-induced fluorescence, Exp. Fluids 29 (2000) 429–437. thermocouple measurements, particularly at lower temperatures. [3] F. Lemoine, Y. Antoine, M. Wolff, M. Lebouche, Simultaneous temperature and The larger (and consistent) discrepancy at denaturation suggests a 2D velocity measurements in a turbulent heated jet using combined laser- systematic source of error that occurs at higher temperature. Inher- induced fluorescence and LDA, Exp. Fluids 26 (1999) 315–323. [4] D. Ross, M. Gaitan, L. Locascio, Temperature measurement in microfluidic ently, the amount of fluorescence change per °C for sulforhodamine systems using a temperature-dependent fluorescent dye, Anal. Chem. 73 B (monosodium salt) is lower at higher temperature. (2001) 4117–4123. When absolute differences between temperature measure- [5] C.T. Wittwer, G.B. Reed, K.M. Ririe, in: K. Mullis, F. Ferre, R. Gibbs (Eds.), The ments and instrument set point temperatures were compared at Polymerase Chain Reaction, Springer-Verlag, Deerfield Beach, FL, 1994, pp. 174–181. denaturation and annealing, the fluorescence-based temperature [6] S. Mondal, V. Venkataraman, In situ monitoring of polymerase extension rate best met target temperatures. Average fluorescence-based and and adaptive feedback control of PCR by using fluorescence measurements, J. set point temperature differences were 0.4 ± 0.2 and 0.4 ± 0.3 °C Biochem. Biophys. Methods 65 (2005) 97–105. [7] H. Kim, S. Dixit, C. Green, G. Faris, Nanodroplet real-time PCR system with laser for annealing and denaturation, respectively. Average thermocou- assisted heating, Opt. Express. 17 (2009) 218–227. ple and set point temperature differences were higher, at [8] P. Lavieille, F. Lemoine, G. Lavergne, M. Lebouché, Evaporating and combusting 0.6 ± 0.3 and 1.1 ± 0.5 °C for annealing and denaturation, respec- droplet temperature measurements using two-color laser-induced fluorescence, Exp. Fluids 21 (2001) 45–55. tively. These results show that fluorescence-based temperature [9] J. Sanchez, C. Phillips, C. Børsting, K. Balogh, M. Bogus, M. Fondevila, C. control can accurately achieve temperature cycling targets, per- Harrison, E. Musgrave-Brown, A. Salas, D. Syndercombe-Court, P. Schneider, A. haps better than thermocouple control. Carracedo, N. Morling, A multiplex assay with 52 single nucleotide polymorphisms for human identification, Electrophoresis 27 (2006) 1713– Because of the sensitive and rapid nature of fluorescence acqui- 1724. sition, there is an opportunity for instruments that maximize tem- [10] L. Sanford, J. Kent, C. Wittwer, Quantum method for fluorescence background perature cycling speed by directly monitoring the sample removal in DNA melting analysis, Anal. Chem. 26 September 2013. [Epub ahead of print]. temperature, not the surrounding environment. Implementation [11] R. Palais, C. Wittwer, Mathematical algorithms for high-resolution DNA of ratio configurations for fluorescence-based temperature should melting analysis, Methods Enzymol. 454 (2009) 323–343. improve the accuracy even further than achieved here. The ability [12] P. Bernard, G. Pritham, C. Wittwer, Color multiplexing hybridization probes to both monitor and control thermal cycling based on the solution using the apolipoprotein E locus as a model system for genotyping, Anal. Biochem. 273 (1999) 221–228. temperature is particularly advantageous for PCR applications with [13] F. Lemoine, M. Wolff, M. Lebouche, Simultaneous concentration and velocity moving samples, such as droplet-based and other microfluidic measurements using combined laser-induced fluorescence and laser Doppler systems. velocimetry: application to turbulent transport, Exp. Fluids 20 (1996) 319– 327. Limitations to the current study include that fluorescence- and [14] J. Sakakibara, K. Hishida, M. Maeda, Vortex structure and heat transfer in the thermocouple-based temperature measurements were made in stagnation region of an impinging plane jet (simultaneous measurements of adjacent (instead of the same) capillaries. The vertical positions velocity and temperature fields by digital particle image velocimetry and laser-induced fluorescence), Int. J. Heat Mass Transfer 40 (1997) 3163– of the optical fiber and thermocouple were aligned between capil- 3176. laries, but are hard to place precisely. The thermal mass introduced [15] M. Bruchhausen, F. Guillard, F. Lemoine, Instantaneous measurement of two- into the solution by the thinner thermocouple and thicker optical dimensional temperature distributions by means of two-color planar laser induced fluorescence (PLIF), Exp. Fluids 38 (2005) 123–131. fiber is not equivalent. In addition, cycling control required that [16] P. Lavieille, A. Delconte, D. Blondel, M. Lebouché, F. Lemoine, Non-intrusive the LightCycler24 be interfaced between two computers (one to temperature measurements using three-color laser-induced fluorescence, Exp. measure fluorescence and send a replacement thermocouple volt- Fluids 36 (2004) 706–716. [17] M. Bruchhausen, A. Delconte, Temperature measurements in polydisperse age to the LightCycler and another for PID control). Ideally, mea- sprays by means of laser-induced fluorescence (LIF) on three spectral bands, surement and control should be combined into a single program. Atomization Spray 16 (2006) 599–614. Nevertheless, we have been able to show temperature control [18] J. Han, K. Burgess, Fluorescent indicators for intracellular pH, Chem. Rev. 110 within 1 °C during rapid PCR of 5.5 s cycles. Successful amplifica- (2010) 2709–2728. [19] C. Estrada-Peréz, Y. Hassan, S. Tan, Experimental characterization of tion was demonstrated by single nucleotide genotyping by high- temperature sensitive dyes for laser induced fluorescence thermometry, Rev. resolution melting of small amplicons. Sci. Instrum. 82 (2011) 074901.  Fluorescence-Based Temperature Control for PCR / L.N. Sanford, C.T. Wittwer / Anal. Biochem. 448 (2014) 75–81 81 [20] J. Sutton, B. Fisher, J. Fleming, A laser-induced fluorescence measurement for [22] V. Natrajan, K. Christensen, Two-color laser-induced fluorescent thermometry aqueous fluid flows with improved temperature sensitivity, Exp. Fluids 45 for microfluidic systems, Meas. Sci. Technol. 20 (2009) 015401. (2008) 869–881. [23] J. Coppeta, C. Rogers, Dual emission laser induced fluorescence for direct [21] H. Kim, K. Kihm, J. Allen, Examination of ratiometric laser induced fluorescence planar scalar behavior measurements, Exp. Fluids 25 (1998) 1–15. thermometry for microscale spatial measurement resolution, Int. J. Heat Mass Transfer 46 (2003) 3967–3974.